Thursday morning I decided to switch sessions, Mogens Norgaard's for Tanel Poder's session. I had to queue up, because the room was fully booked, but not everybody showed up, so all people in the line-up could enter the room. I sure was glad I switched, as this was the best session at this conference for me. The title was "Advanced Oracle Troubleshooting: No magic is needed - a systematic approach will do". He outlined a systematic approach to answer the question "What the $#*&%! is that session doing?". My personal favourite SQL*Trace collects all work done for one process or single query, but this isn't any good when the process or query doesn't return results. And that's exactly what this method was about. Tanel showed his collection of scripts that makes the querying against v$session_wait, v$session_event and v$sesstat a lot easier. And when that doesn't help, there's always the process stack of course :-). And actually it is a lot easier than it looks initially, especially with Tanel showing it. All qer... calls are just query execution row source lines, corresponding to row source operations you see in your execution plan. The meaning of each individual line can be looked at MetaLink note 175982.1. A lot of eye-openers in an awesome session. You can find the material he used here on his blog.
Then I went to see the "Oracle ACE Director Birds-of-a-Feather Tips and Techniques Panel" by Oracle Aces Lewis Cunningham, Arup Nanda, Tim Hall, Eddie Awad, Mark Rittman, Hans Forbrich and Bradley Brown. All attendees could write a question on a piece of paper, the papers were collected, Lewis read the questions and the panel answered them. Didn’t hear anything new, but it was entertaining.
Next session was Juan Loaiza session called "Oracle Exadata Architecture Overview". He explained the techniques behind the Exadata concept in more detail, even the physical layout. In short it ships less data through pipes by using smart scans (see this whitepaper), has more pipes and has bigger pipes using Infiniband. It is designed to gracefully tolerate cell brownouts. Not really relevant to an average developer, but the concept of smart scanning can be relevant when trying to understand how queries are executed in this architecture.
Then Chen Shapira’s unconference session about Streams. She explained the Streams concepts Capture, Propagate and Apply. In the OTN Streams Forum she participates in, a lot of questions are like "I configured Streams exactly like it said in the documentation, but it doesn’t work. Help!" and that’s why she spent the main part explaining how to troubleshoot, using a live demo. It’s about systematically checking all parts. Initially the source must be equal to the target. Start with Capture; check dba_capture.status to see if it is enabled or look at the alert.log to see if anything is wrong. For Apply, check the automatically created error queue. And then check the rules. I entered the session only knowing Streams is something about replication with redo logs using Advanced Queueing under the covers, and now I think I actually understand the whole process conceptually and I know where to look when something doesn’t work. So the session was very good. Check out her blog for more details on Streams and this session.
The last session of Oracle OpenWorld was Gaya Vaidyanatha’s "Oracle Performance Management: Real-World Case Studies". He’s the guy who invented the term Compulsive Tuning Disorder, explained in the "Oracle Insights: Tales of the OakTable" book. The session was a collection of war stories where the developer – of course – did something extremely stupid, how he detected the problem in the trace files and what the solution was and how much it solved. And how he made many hardware vendors extremely pissed off with his advice. Gaya is an excellent speaker: strong voice, charismatic and humorous. So this was an entertaining end of Oracle OpenWorld.
Looking back at Oracle OpenWorld I think it was a fantastic event. It’s a great place to meet like-minded people you know from the articles, blogs and OTN forums. There were a lot of very good sessions, but you have to choose them carefully or you'll end up in some marketing talk. Also compliments to the team who organized the event. I cannot fathom the work that needs to be done to make this all happen, but they did. And of course, thanks to my manager Roel, for allowing me to visit Oracle OpenWorld! Next year again?
Sunday, September 28, 2008
Oracle OpenWorld, day 5
Posted by Rob van Wijk at 1:53 AM
Subscribe to:
Post Comments (Atom)
Rob hi,
ReplyDeleteIt was a pleasure meeting you, I hope we meet again soon, maybe this december at UKOUG? :)
I wrote an email to you, I hope you received it?
Warm regards from Istanbul,
Tonguç
Hi Tonguç,
ReplyDeleteIt was very nice meeting you too. UKOUG looks very interesting with good speakers, but I don't want to push my luck so I won't be there unfortunately.
And yes, I received the e-mail. You've got one back :-)
Regards from Utrecht,
Rob.